[Daniel's week] May 24, 2024

[Daniel Stenberg]

Hello,

Another packed week ends...

## curl talk

I took the train to Gothenburg on Monday and talked curl to the good people at 
Mullvad VPN. A super attentive audience with a lot of quality questions. The 
amount of general appreciation and yeah, I guess love, for my work and 
accomplishments was also at a high level.

A reminder to myself that it is fun and sometimes very constructive to leave 
home every once in a while and communicate with real people.

I did a quick count and it turns out I have done on average 15 public talks 
per year the last nine years.

## curl 8.8.0

The talk activity distracted me a little bit so I was behind of the final pull
request merges before the release, which made me do it a little too many of
those on late Tuesday which of course is a little too late for everything to
feel entirely safe and comfy.

Still, by 8am on the Wednesday I announced curl 8.8.0 to the world [1]. The
first regression was mentioned after 40 minutes, and after forty-five minutes
I had a fix for that in a fresh pull request. It was not a particularly large
problem so this one alone is not going to make us do a patch release.

I did my usual live-streamed release presentation at 10, which this time
expanded and went on a little longer than usual. Mostly because there was such
a fine conversation going and it lead me into creating yet another graph for
the dashboard: average and median number of lines in the source code for the
library and for the curl tool. You can now find it updated daily in the
dashboard [5] among the others.

There was a new record broken in this release: we have never shipped this many
bugfixes in a single release before: 220.

## reproducible

Since we have worked a bit on making the curl releases reproducible lately,
with the latest enhancements shipped in 8.8.0 I figured I would tell everyone
exactly how such a reproduction and verification can be done now. I wrote a
blog post about it[2].

## colon slash slash

Someone asked me the other day about the :// in the curl logo and what I
thought about Mozilla's use of it. It was then a fun coincidence that Frank
Gavaerts this week brought back some logo sketches he and I worked on back in
2015 and that was enough to trigger me into putting all those details into a
blog post to explain how the current curl logo came to be [3].

## graphs

After the graph work I did live on the stream I also took the opportunity and
fixed up some other graphs. Some of the more recent ones I made that compare
different things with the amount of production lines of code were perhaps not
done in the most clear way as I have been receiving questions about them. I
originally made them in the style of "number of lines of code per test case"
which then makes it a falling graph when things are good. When there are less
code per test case.

I have now reversed that graph and a few others to instead be "number of test
cases per KLOC" where KLOC of course is 1,000 lines of code. This seems to be
easier to grasp and it makes the plot climb on the right side when things are
going in the right direction.

The dashboard [5] now shows that we have received more pull requests in May
2024 than we have done any other month in curl's history. With a week left of
the month.

## c-ares 1.29.0

We shipped another c-ares release today [6]. A few new features and a bunch of
bugfixes. Check it out!

## survey

The curl user survey [7] is still running for a few more days. I will take it
down end of day Tuesday. We have now received an amazing 1,300 responses but
let's hope we manage to a get a few more before this ends.

## peer bonus

Aka beer bonus. I had the honor to receive beer money from Google again. For
the fifth time[4].

## curl work

In near term I will work on adding support for CA caching when using wolfSSL,
in the same style as we already support for OpenSSL. Ideally we do it for
GnuTLS more or less at the same time. CA caching can provide a significant
performance boost for some users.

We have pending work on a way to improve how curl deals with closing TLS
connections and by this reduce the risk that we skip the close-notify TLS
take-down part and instead of closing down the TCP connection nicely end up
with a RST and getting the connection in a wait state.

I intend to make my first test shots at packaging curl for QNX. I can build,
but exactly how the packaging should be done is still not crystal clear to me
but I figure we can back-and-forth this a little once I provide my first real
attempts.

## coming up

- Tuesday: the curl user survey ends
- Wednesday: talk curl to open source peeps at the European Commission
- Thursday: talk HTTP/3 at Webbhuset, Gothenburg [8]

## links

[1] = https://daniel.haxx.se/blog/2024/05/22/curl-8-8-0/
[2] = https://daniel.haxx.se/blog/2024/05/23/how-to-verify-a-curl-release/
[3] = https://daniel.haxx.se/blog/2024/05/21/a-history-of-a-logo-with-a-colon-and-two-slashes/
[4] = https://daniel.haxx.se/blog/2024/05/24/google-peer-bonus-number-five/
[5] = https://curl.se/dashboard.html
[6] = https://c-ares.org/
[7] = https://daniel.haxx.se/blog/2024/05/14/curl-user-survey-2024/
[8] = https://www.meetup.com/webbhuset/events/300590515/


-- 

  / daniel.haxx.se